Your brain has an age, and a blood sample can guess it
A Stanford team read eleven organ ages from 44,498 blood samples. The brain and immune system told the most. What that predicts is narrower, and better, than the headlines.
Theme
Science & Progress
Medicine, energy, space and AI as engines of a longer, freer life, and what each new result really changes.
Filter: Essays · Long reads · Daily Pick · Search inside
Biological Age & ClocksMuscle & Body CompositionSpaceBrain & MindHeart & LipidsGLP-1 & Metabolic HealthPipelines & TrialsPlanet & EnergyReading the EvidenceWearables & Body RhythmsAI Models & BenchmarksCorrections & Retractions
84 essays
The Magnet That Wouldn’t Flatter Us
“We say we want revolutions. What we need, more often, is a measurement that refuses to flatter us.”
The GLP-1 muscle scare is half wrong, and the other half is up to you
About a quarter of the weight lost on GLP-1 drugs is "lean mass." That is not the same as a quarter being muscle. But the drugs also shrink what people eat, protein included, and the only proven defenses are the oldest ones: enough protein and a barbell.
Perseverance broke the off-world driving record by letting the rover decide
NASA's Mars rover passed Opportunity's 28.06 miles in five and a half years instead of fifteen. The odometer is the headline; the real record is how much of the driving the machine now does on its own judgment.
In aging brains, more myelin tracked faster decline. The sample was small
A Nature Medicine study of the Lothian Birth Cohort links steeper cognitive decline after 70 to thicker myelin and fewer large nerve fibers, and points to a protein called NRF2. It is a striking idea, built on a dozen or two brains.
For heart risk, metabolic health often counts more than weight. Weight still counts
A meta-analysis of 44 cohort studies finds that thin people with poor metabolic health often fare worse than heavy people with good metabolic health. Its own conclusion is more careful than its title.
Six patients, two years, and a Phase 3 of eight: how gene therapy for LCA5 blindness is being tested
Opus Genetics says vision gains in three adults treated for LCA5 held through 24 months. The numbers are tiny and come from the company, but the regulatory path built around them is the more interesting story.
A weaker drug you can take at home is still a real choice
The FDA has approved Genentech's Enspryng for thyroid eye disease, the first treatment patients can inject themselves. One of its two pivotal trials missed its goal, and its effect looks smaller than the infused incumbent's. The approval is still right, and the reason why matters.
Thwaites, three centimeters, and the physics models left out
A new model of Antarctica's Thwaites Glacier that lets seawater seep under the ice projects up to 30 millimeters of sea-level rise by 2100, about twice the standard ensemble mean, even with no further warming. The doubling depends on one parameter, and that is the interesting part.
Eight old mice, a peptide and a treadmill. The 42 percent is real, the drug story isn't finished
A Rowan University team gave aged mice an autophagy peptide on top of running. Grip strength rose 42 percent in the combined group. Read what that number compares, and what the muscle samples failed to show.
The half-million deaths we already know how to prevent
A model presented at an American Heart Association meeting says better blood pressure control in 20 countries could prevent 518,476 cardiovascular deaths a year. The six-digit precision is spurious. The underlying point is not, and it is not a medical problem.
Your watch can tell you to see a sleep doctor. That is already a lot
A Stanford study found a Samsung Galaxy Watch caught 94 percent of moderate-to-severe sleep apnea in high-risk adults. It also flagged about a third of people who didn't have it. Both numbers matter, and one press line got them wrong.
Nineteen old mice, one splicing gene, and a word worth reading twice
A Harvard-led team says boosting SRSF1 made old cells look younger and was associated with 14 percent longer median life in mice. The result is real and elegant. The sample, and the gene's past, call for patience.
Europe said yes to Lilly's GLP-1 pill. Patients will hear it country by country
The EMA's committee backed Foundayo, an oral GLP-1 tablet, on October 15. That is a scientific verdict, not access. The Commission still has to sign, and then every member state decides what it will pay.
Apple pulled its new Health app from the beta. That was the plan
iOS 27.2 beta 4 removed the redesigned Health app, added switches for Messages suggestions and drained the red from Apple Music. The most telling line sits in the developer notes, under App Tracking Transparency.
ApoB counts the particles. Not everyone needs the recount
When apoB runs high while LDL looks fine, arteries tend to calcify faster. That is real. So is the case, made this fall by preventive cardiologists, for testing selectively rather than for everyone.
Older adults who nap most days face more dementia later. The nap is probably the messenger
A Johns Hopkins analysis of 995 Medicare beneficiaries links frequent and unplanned napping to a 48 percent higher hazard of possible or probable dementia over ten years. Before anyone bans the siesta, read what was actually measured.
Twenty-one players, six injuries, and a blood score worth testing properly
A German study says a six-point blood score beat wearable workload data at spotting which footballers would get hurt within ten days. The idea is sound. The evidence is one squad, one summer, and thresholds fitted to the very players they predict.
In Parkinson's, the hours seem to matter more than the sweat
A seven-year Cincinnati cohort links more than eight hours of weekly activity to slower motor and cognitive decline. The most interesting finding is not the threshold. It is that intensity stopped mattering once total time was counted.
Metformin lengthened telomeres by 29.6 percent in six months. That is the problem
A small Indian trial reports a huge jump in a classic marker of aging in people with prediabetes. The authors themselves urge caution. They are right to, and the trial registry gives more reasons.
A free liver score keeps flagging heart risk in diabetes. Flagging is not predicting
Three new studies, from Vietnam, Hungary and a cohort of 318,308 people, link the FIB-4 index to cardiovascular trouble. The signal is consistent. What it adds to a doctor's judgment is still small.
A sleep apnea implant for patients who had none. Strong numbers, no control group
Nyxoah's Genio cut breathing interruptions by about three quarters in a group of sleep apnea patients who are currently barred from nerve-stimulation implants in the US. The result is promising and the need is real. But 47 patients, no comparison arm and a company-run readout mean the FDA, not the press release, should have the last word.
"Protein added 7 kg to the squat. The lifting did most of the work"
A new meta-analysis finds that protein supplements on top of strength training add a few kilos to the bench and squat in young adults. The fine print says the effect depends on how long you train and how much protein you already eat. The barbell matters more than the shaker.
Two viruses, one protein: a clever fix for dysferlinopathy, still in mice
A gene too big for one viral vector gets split in two and stitched back together inside the muscle cell. The preclinical results are real and the dose story is interesting. But the paper was written largely by the company's own founders, and no human has received it yet.
A blood metabolite clock built on 18,476 people, ages 1 to 96. Useful, if you read the fine print
Chinese researchers used 56 plasma metabolites to estimate "metabolic age" and found that running ahead of the calendar tracked fatty liver, hypertension, diabetes and death. The cohort is huge. What we can learn from the coverage is narrower than the headlines.
"Same family, opposite jobs: two fat-cutting enzymes and the aging memory"
In one week, two mouse studies found one phospholipase that protects the aging hippocampus and another that wears it down. The lesson is less about a pill than about how specific the brain's bookkeeping is.
An Alzheimer's blood test reads lower in some groups. Does it still read true?
In 15,612 people screened for a prevention trial, Asian, Black and Hispanic participants had lower levels of the blood marker %p-tau217. Its link to brain amyloid held across racial groups. Across ethnicity, it did not quite.
The first mitochondria drug turns one, and the bill comes due in 2029
Forzinity brought in $17.5 million gross in its first ten months on the market, for a disease most people have never heard of. The number is less interesting than what it says about how rare-disease medicine actually gets paid for.
A glucose sensor did what a new pill would brag about
In a randomized trial, people with type 2 diabetes who don't take insulin cut their A1C almost a point more with a sensor than with routine care. The result is real. So is the fine print.
The Congo is losing its forest the way poor countries lose everything
Deforestation fell 15 percent in 2025, and tropical Africa now clears almost as much forest as tropical Latin America. The difference between the two is not concern for nature. It is a state that enforces, and an economy that offers something better than charcoal.
A gut bacterium makes its own statin. In mice, it halved the plaque
Researchers in Shenzhen traced a heart-protective effect to a fatty acid made by Bacteroides uniformis. The mouse results are striking. The peer review file, published alongside, is the best reason to keep reading carefully.
A weight-loss pill that bets on being gentle
Biomea Fusion's oral GLP-1, BMF-650, produced modest weight loss in four weeks and unusually mild side effects. The interesting part is the strategy, not the number, and the market did not buy it yet.
Two visual puzzles spot early trouble in Lewy body disease. They do not diagnose it yet
Queensland researchers built tests hard enough that no healthy person aces them, and found visual deficits in about half of Parkinson's patients with no cognitive complaints. The finding is real; the "diagnostic test" headline is ahead of the data.
Every 10 mmHg lower cut heart events 15 percent, kidney disease or not. The kidney numbers deserve a careful read
An Oxford-led meta-analysis of 27 trials and 88,830 people finds intensive blood pressure control protects the heart just as well in chronic kidney disease. It also finds more drops in kidney function, without more kidney failure.
A pancreatic tumor is not one enemy. It is a crowd that keeps changing its mind
An NYU-led team read 443,451 single cells from 62 pancreatic tumors and caught clones switching to their most dangerous form. One gene, MYC, keeps showing up at the switch. That is a map, not yet a cure.
Claude Solved a Crypto Mystery and Made Quantum Attacks Cheaper
The NIST curve seeds were just 1990s timestamps. The same day, a new quantum circuit used 17 times fewer gates. Bitcoin's real test is governance.
Two elegant new theories of why old muscles fail, and the barbell still wins
One team blames a fading sodium channel at the nerve-muscle junction; another blames a tired liver that lets lactic acid pile up in muscle. Both are beautiful. Both rest mostly on rodents, while the tools with human trials behind them are stubbornly unglamorous.
Year two of a Duchenne cell therapy, and the limits of watching
Boys with Duchenne who switched from placebo to deramiocel lost arm function far more slowly in their second year. That is encouraging, and it is exactly the kind of evidence that can't settle the question the FDA must answer by November 22.
Isembyld is a real first for SMA. Its label tells you exactly how much of one
The FDA approved the first muscle-targeted drug for spinal muscular atrophy in September. The prescribing information is more candid than the launch: a modest gain in children, an adult indication built on modeling, a fracture warning, and a list price near $455,000 a year.
Two in five children left the obesity range. Now show us the rest
Novo Nordisk's STEP Young trial gave semaglutide to children aged 6 to 11, and 40.4 percent no longer met the definition of obesity after 68 weeks. The result deserves hope, not panic. It also deserves the numbers the press release left out.
Groceries lowered blood sugar in a real trial. More groceries did not lower it more
A randomized trial in Medicaid patients with type 2 diabetes found that weekly deliveries of healthy food cut HbA1c by 0.40 points more than usual care. That is a real result. But the bigger box did no better than the smaller one, almost nobody used the counseling, and the question of cost was never measured.
Roche's diabetes numbers are striking. The placebo arm is missing
Enicepatide lowered HbA1c by 2.65 points and weight by 15.5 percent in a 48-week trial, says Roche. The release never says what happened on placebo, and the public registry names a different week as the primary endpoint. Both details matter more than the headline.
An NYU team found a conductor for muscle repair. So far it plays for mice
Extra doses of one RNA-binding protein, AUF1, kept severely injured mouse muscle from wasting and preserved its strength. The idea is elegant, the company behind it is real, and the distance to a human knee is still long.
A capsule in the pocket for a disease that swells without warning
In a phase 3 crossover trial, oral deucrictibant brought first relief from hereditary angioedema attacks in a median of 1.28 hours, against more than 12 on placebo. It would not be the first pill for attacks, but it may be a better kind of choice.
Thirty-seven minutes, and a new door into Parkinson's
Cerevance's solengepras cut daily OFF time by about 37 minutes more than placebo without touching dopamine receptors. The effect is modest. The mechanism is what makes it worth watching, and the unreported endpoints are what make me wait.
Seven FDA decisions before Christmas, and seven different kinds of proof
Between late October and late December, the FDA is due to rule on drugs for hepatitis B, lung cancer, Duchenne, kidney disease, breast cancer, myeloma and obesity. Read the trials behind them and you get a short course in what counts as evidence.
Where your arm rests can turn a normal reading into a diagnosis
A Johns Hopkins crossover trial found that an arm hanging at the side added 6.5 points to systolic blood pressure, and nearly 9 in people already running high. The fix costs nothing. Ignoring it can cost a prescription.
Semaglutide protects the heart. Nobody can yet say exactly how
A new analysis of the SELECT trial could not pin the drug's 20 percent cut in heart attacks, strokes and cardiovascular deaths on weight loss or any other measured risk factor. That is a more interesting result than the headline "not just weight loss," and a humbler one.
The weight-loss pill that ignored menopause, mostly
Lilly's oral GLP-1 produced similar weight loss in women before, during and after menopause. The good news is solid in one trial and thin in the other, and the difference is worth reading.
One insulin shot a week, and the fine print that comes with it
The FDA has approved Lilly's Onswik, a once-weekly basal insulin for adults with type 2 diabetes. It matches daily insulin on blood sugar. Its real promise is convenience, and its limits are written into the label.
The artificial pancreas is becoming a set of parts you can choose
A new patch pump cleared, an implanted year-long sensor joining an automated system, and a year of Omnipod 5 data from France. The real story is not any one device. It is that people with type 1 diabetes are starting to assemble their own.
Eight hours or nine is the wrong question about eating windows
A pooled analysis says a shorter window means a little more weight loss. A 247-person trial says nine hours did about as well as a dietitian, and neither moved blood sugar. Read together, they say the clock is a tool for discipline, not a metabolic switch.
Twelve weeks of shakes and squats put half of young diabetics into remission. Now comes the hard part
In the RESET for Remission trial, 54% of adults under 46 with early type 2 diabetes were off medication with normal-range blood sugar after six months, against 4% on usual care. The result is striking. It is also small, short, and honest about what it cannot prove.
The digital slime mold, and the $55 boxes it would live in
Anthropic ran a 26-billion-parameter model on fifteen cheap Android TV boxes and tested whether small AIs can copy themselves. They showed less than the word "worm" suggests, and the weak link is older than AI.
The seizure monitor that asked "Was that one real?"
A Penn team let thirteen epilepsy patients answer their EEG through a chatbot, then retrained the detector on those answers. In a replay, false alarms fell by about three quarters. The idea is bigger than the numbers, and the numbers are smaller than the idea.
Four pairs of eyes, 46,000 miles out, and a Moon that is not gray
NASA's preliminary science report from Artemis II reads like a field notebook. Five faint impact flashes, a "jaundiced cheek," mountains on the limb. It is the best argument in years for sending people instead of only cameras.
The most valuable thing a machine writes is the part you never see
Anthropic now summarizes, fences and polices Claude's reasoning to stop rivals from copying it. The company says it cut what distillers harvest by 90 to 99 percent. The harder question is who pays for the fence.
Twenty years, one arm, one booster: what the brain-training trial really shows
A trial begun in 1998 links a few weeks of speed training to fewer dementia diagnoses two decades later. The result is real, narrower than the press release, and now the subject of a sharp scientific quarrel worth following.
K2-18b and the life signal that depended on how you sliced the data
In 2025 a Cambridge team reported the "strongest hints yet" of life beyond the solar system. An independent reanalysis found the signal appeared in only one of eight ways of binning the same data. The episode is less a scandal than a demonstration of science working, slowly and in public.
The ocean's carbon pump depends on how sticky its snow is
A Scripps team spent four years learning to see bacteria inside marine snow. The finding, that gooier particles resist decay and sink deeper, is a lovely piece of basic science. The "six times more carbon" figure comes from simulations, and that distinction matters.
Is antimatter leaking out of the Milky Way? The authors are more careful than the headlines
A paper in Astronomy & Astrophysics links faint gamma-ray hotspots to gas clouds and nearby galaxies, and suggests our galaxy makes more antimatter than we thought. Read the significance first, and the paper's own caveats second. Then wait for the telescope launching next year.
Google's nuclear deal builds no reactor, and that is its genius
Google will pay for 890 megawatts of new nuclear output by upgrading eleven reactors that already exist. It is the fastest clean power on offer. It is also a well that will run dry.
How many gentoo penguins are there? Counting is a policy choice
A genomic study of 64 gentoos across the Southern Ocean argues the "single species" is really four, and names a new one from an 1897 museum skin. The biology is persuasive. What makes it matter is that a "Least Concern" label may be hiding three lineages in trouble.
Strokes Before 55 Nearly Doubled. Heart Attacks Went the Other Way
Cincinnati counted 2,076 first strokes in adults under 55: the rate rose from 34 to 62 per 100,000. In Oxford, young heart attacks fell over the same years.
Pelacarsen lowered Lp(a) and did not prevent heart attacks. That is the lesson
The first outcomes trial of an Lp(a)-lowering drug, in more than 8,000 patients, missed its primary endpoint. The marker moved; the events did not. Testing still matters, and the next trials just became more important.
September's most useful AI skill was knowing when to stop
OpenAI halved its prices, Anthropic shipped a Sonnet that nearly matches its flagship, and the independent tests complicated both stories. The most instructive model of the month was the one that never shipped.
Helping Others Slowed Cognitive Decline in People Carrying the Alzheimer's Risk Gene
A free behavior, a few hours a week, and a slower slope of decline for the gene that usually steepens it. The data are observational, and the paper says so.
A PCSK9 pill, at last. The proof that it saves lives is still borrowed from a shot
Merck's enlicitide cuts LDL cholesterol by more than half in a daily tablet. What it has not yet shown is fewer heart attacks, and a fight over which number to chase makes that gap matter more.
The ring on your finger is guessing, and it guesses best when you need it least
Sleep trackers are decent at telling sleep from wake in healthy young adults. Ask them about deep sleep, older sleepers or patients, and the numbers drift. A lawsuit over Oura's "95%" shows why the label matters as much as the sensor.
Retatrutide takes weight off, Viking tests a monthly shot to keep it off, and the fine print decides how impressed to be
Lilly's triple agonist cut weight by up to 18.8 percent in people with diabetes, or 20.8 percent, depending on which estimate you read. Viking's maintenance data are promising but small and short. The obesity race is moving from "how much" to "how long," and the numbers need careful reading.
The lonely waves of an aging brain
A Berkeley study found that frontal tau goes with sleep waves that stop traveling, and with weaker overnight memory. A Yale study linked less deep and REM sleep to a smaller Alzheimer's-vulnerable brain region years later. Both are elegant. Neither tells you yet whether better sleep would protect you.
Your aging clock responds to a salad and a run, and to prescription drugs even more
A Yale analysis of 51 intervention studies found that diet and exercise nudged epigenetic clocks, supplements did not, and drugs did most. The finding is useful. So is noticing who sells the clocks.
In Tanzania, the AI's best trick was giving nurses their afternoon back
The win was not a benchmark score. It was people with sight threatening eye disease actually reaching a specialist, and reaching one weeks sooner.
The blood pressure pill whose real number is 3.8
Against placebo, the endothelin blocker lowered systolic pressure by about 4 mmHg on top of three drugs. New post hoc data on obesity and kidney risk.
One clock is a rumor. Three clocks are a lead
A Yale team reran published "age reversal" results on eleven extra epigenetic clocks and none of them agreed. The same week, a skin cream, a weight-loss drug and a diabetes drug all claimed to turn back biological time. Here is how to tell which claims deserve a second look.
Two amylin drugs "beat" tirzepatide. Look at which tirzepatide
Novo's CagriSema topped tirzepatide 5 mg, the low end of the range. Lilly's EloraTZP topped tirzepatide 15 mg, the top of it, in a small trial where many people quit. Both results are real. Neither means what the headline wants.
Myostatin blockers cut GLP-1 muscle loss. Nobody has shown they make you stronger
Two antibody trials now show that blocking myostatin can roughly halve the lean mass people lose on semaglutide or tirzepatide. That is real progress. But the one trial that measured physical function found no difference, and an add-on drug sold on "quality of weight loss" will have to prove it means quality of life.
The Lp(a) Test Everyone Should Get Once, and What It Can't Tell You
One blood test for life sounds simple, but this year's research shows the result can move, the risk has no clean cutoff, and no Lp(a) drug has proven it saves hearts.
Older adults built muscle without meat. They built less of it
Twelve weeks of lifting: vegetarians gained 0.38 kg of limb muscle, omnivores 0.66 kg. The trial could not show the two diets were equivalent.
Meta reached the frontier and closed the door behind it
Meta's new flagship scores 56 on an independent index, two points behind Claude Opus 5.5. And the stronger model is not the one you can call today.
Lilly's GLP-1 pill passed its heart-safety test. The death numbers are a question, not an answer
ACHIEVE-4 showed that oral orforglipron is no worse for the heart than insulin glargine, which is what it was built to show. The striking drop in deaths is real data and an unproven claim at the same time.
"The pill duel: orforglipron beat oral semaglutide, and the fine print matters"
The first head-to-head trial of two oral GLP-1 drugs, published in The Lancet, gives Lilly's pill a clear win on blood sugar and weight. It also shows more stomach trouble and more people quitting. And the press release led with the friendlier of two sets of numbers.
A pill to protect muscle on semaglutide reaches Phase 2. Its main test is fat
Sironax says its NAMPT activator could help people lose fat on GLP-1 drugs while keeping lean mass. The trial is halfway enrolled, the evidence so far is from mice, and the registry quietly reorders the promise.
Your wrist keeps a diary of your days
A week of light and skin-temperature readings from 809 Spanish adults lined up with their blood fats, insulin and, in women, their body fat. It is a lovely finding and a cautious one, and the two belong together.
The Sunday letter
One email on Sunday: the essay of the week and the question worth carrying into the next one, from both sides of the Atlantic. The letter is paused for now: no signups and no sends while the site is rebuilt.
Signups are paused while the site is rebuilt. No letter is being sent.